Classification of Drug Hypersensitivity into Allergic, p-i, and Pseudo-Allergic Forms

Adverse drug reactions (ADR) are common. They are due to different mechanisms and result in different clinical pictures. In 1977 and 1981, Rawlins and Thompson [1 ]proposed a subclassification of ADR, which is still widely used today for an initial approach to ADR [1,2]: type A reactions are due to the pharmacological activity of the drug. They are influenced by drug pharmacokinetics, comorbidities and/or drug-drug interactions. Overdosing and drug binding to off-target receptors are central to this “pharmacological” reaction type. Type A reactions occur in nearly all individuals, are dose dependent, are considered to be predictable, and their mechanism is mostly understood (Table 1). A typical example would be sleepiness caused by first-generation antihistamines.

Type B reactions comprise about 15% of all ADR. They occur in individuals with a certain predisposition, and were called “idiosyncratic” ADR. They are not readily anticipated and thus were, from a pharmacological perspective, called “bizarre” ADR (Table 1) [3]. A few are due to certain enzyme deficiencies and may appear as a nonimmune-mediated “hypersensitivity” reaction to drugs but should be better classified as exaggerated reactions - to distinguish them from classical immune-mediated hypersensitivities. An example is hemolysis after drugs (antimalarials like primaquine, certain analgesics, and antibiotics) in individuals with glucose-6-phosphate dehydrogenase deficiency [4].

The majority of type B reactions involve the immune system and are drug hypersensitivity reactions (DHR). We will not use the term drug allergy, as this refers to a specific immune response to a drug acting as an allergen (mostly linked to a protein or peptide), while drug hypersensitivity is a term going beyond drug allergy: immune stimulations and corresponding clinical symptoms can also occur when drugs bind directly to immune receptors like HLA or TCR (T-cell receptor) proteins (pharmacological interaction with immune receptor[p-i] concept) [5,6,7] or when inflammatory cells are stimulated by drug-receptor or drug-enzyme interactions (pseudo-allergy) [8,9]. Thus, the need of covalent linkage of a drug to a protein carrier molecule to form a new antigen is no more considered a prerequisite for generating a DHR [5,6,7,8,9]. Indeed, a substantial part of DHR follow pharmacological but not immunological rules and actually show features of a type A reaction. These new findings have an impact on clinical course, management, prediction, dose effect, and cross-reactivity of such DHR, and thus should be considered in the clinical approach to a patient with DHR.

ADR to a therapeutic antibodies or other proteins/peptides follow distinct rules and need a special approach [10]. Later, additional forms of ADR were described (ADR type C-E), but these subclassifications are rarely used [10]. Recently, ADR under therapy with checkpoint inhibitors were distinguished from other ADR and called immune-related adverse reactions. Some of these immune-related adverse reactions may be related to loss of normal control mechanisms by the immune system due to therapy with anti-PD-1 or -CTLA-4 antibodies [11].

The strengths of the type A and B classification is its simplicity and the clear definition of type A reactions with its associated features, whereas type B reactions remain less clear and are essentially defined as “non-A” type (Table 1). What did not fit into type A was classified as type B [1,2,3]; ADR type B were not due to pharmacological actions, were not predictable, not dose dependent, and result in a nonrational, “bizarre” clinical picture. This “non-A” classification had a substantial negative impact on the clinical management and research of type B reactions. As type B reactions were thought to be due to idiosyncratic, individual features of the patient, idiosyncrasy was interpreted as“the drug itself is OK, just the patient is weird.” Type B reactions were considered to be rare but unavoidable. The claim of dose independence ruled out dose reduction as a therapeutic or prophylactic option. No animal models were developed, or if they existed in pharmaceutical industry, they were not analyzed in detail as the drug development was simply stopped if reactions occurred.

In spite of these obstacles, several academic centers tried to improve the understanding of the clinical features and pathomechanisms of DHR. By studying immune-mediated immediate- and delayed-appearing hypersensitivity reactions in the affected patients, it was found that the interaction of a drug with the immune system was more complex than previously anticipated [7].

Different attempts have been undertaken to subclassify DHR, with the aim to better diagnose, manage, and possibly avoid them (Table 2). In clinical practice, these approaches are often combined.

- The time until symptoms appear (within 1 h or thereafter) is the simplest, useful (for some reactions), and still widely used classification [12]. However, it does not consider the characteristics of the drug, type of the immune response, dose dependence, and individual susceptibility.

- An immune-mediated mechanism linked to certain clinical phenotypes is the basis for the (revised) Coombs and Gell [13 ]classification [14]. The immediate-appearing symptoms (urticaria, anaphylaxis) were classified as being due to IgE and mast cell degranulation, and the delayed-appearing symptoms (exanthems, hepatitis) as dependent on T-cell activation (and rarely antibody involvement, especially IgG). This classification and attribution of symptoms to an underlying immune mechanism are also important to devise optimal testing strategies, as the type of immune stimulation determines the procedures needed for skin or in vitro testing with the incriminated drug.

- Some reactions known to occur mainly with certain drugs were named accordingly, e.g. NSAID intolerance, anticonvulsant hypersensitivity syndrome, or acute infusion reactions to biologicals.

None of these classifications is ideal and per se able to explain and link pathomechanisms, clinical pictures, dose dependency, cross-reactivity, and optimal treatments (Table 2). Moreover, all classifications assumed that the formation of a new antigen by the hapten-carrier concept is essential for DHR. Here, we propose to also consider direct, noncovalent drug-receptor interactions as cause of DHR and to differentiate three major forms of DHR: allergic/immune, pseudo-allergic, and pharmacological stimulations.

This mechanism is the classical explanation for DHR and represents a true drug allergy. It is based on the data generated in the early 20th century [15] and states that small molecules like drugs or drug metabolites are too small to elicit a specific immune response on their own. Only if they bind covalently to proteins a new antigen is generated (hapten-protein complex) [16]. Some drugs are prohaptens and gain the ability to bind covalently to proteins only after being metabolized [17]. Table 3 shows a list of drugs acting as haptens. The stable covalent binding is important for two features of haptens:

- The ability to stimulate cells of innate immunity (e.g. dendritic cells) [18,19]: the drug activates pattern recognition receptors - either by direct chemical interaction with such receptors or by induction of endogenous activators. This induces the expression of CD40 or other costimulatory molecules on dendritic cells [18,19].

- The covalent binding by a drug alters the protein structure and can transform an autologous soluble (e.g. albumin, transferrin) or cell-bound protein (e.g. integrins, selectins) to a novel drug-modified protein, which acts as a new antigen [16,20,21] to which no intrinsic tolerance exists. The adaptive immune response reacts via specific receptors (TCR for antigen on T cells and immunoglobulin receptor on B cells) with these newly formed antigens [16,20].

The covalent link of a hapten to its protein carrier is resistant to intracellular processing. Hapten-modified peptides are finally presented on HLA molecules to TCR [16,21]. In most instances, haptens bind to a specific amino acid (e.g. lysine) in different positions within the protein. Thus, after protein processing to smaller peptides, various hapten-peptide conjugates are generated which bind to different HLA molecules [16,20,22]. The resulting immune responses to hapten-protein complexes is therefore not restricted to a single HLA molecule (in contrast to some p-i reactions, see below).

In conclusion, a hapten elicits both T- and B-cell reactions simultaneously and is normally not linked to a single HLA allele. Hapten reactions are polyclonal and functionally heterogeneous, involving antibodies, cytotoxicity, and cytokines released by T cells and activated inflammatory cells [14]. They appear rapidly if the effector mechanism relies on cell-bound IgE crosslinked by hapten-protein complexes and subsequent mast cell/basophil degranulation. On the other hand, they might appear delayed if the effector mechanism relies on prior expansion of T cells stimulated by hapten-peptide- HLA complexes followed by recruitment of inflammatory effector cells.

The concept that drugs are too small to represent full antigens is correct. However, drugs designed to fit into “pockets” of certain protein structures may also directly bind to immune receptors like HLA or TCR proteins [5,6,7]. This drug binding to immune receptors is a typical off-target effect of the drug and is analogous to the binding of drugs to other receptor structures: it is based on noncovalent bonds like van der Waals forces, hydrogen bonds, and electrostatic interactions. Of note, the drug does not bind to the immunogenic peptides presented by HLA. Thus, the drug is not acting as an antigen or substituting an antigenic peptide. This reaction type is termed p harmacological interaction of drugs with i mmune receptor (p-i) [5].

The targeted immune receptors are highly polymorphic, both in each individual (TCR) as well as in the population (HLA): one estimates that there are >10¹¹ different TCR per individual and >10,000 different HLA class I and >3,000 class II alleles in the human population (www.allelefrequencies.net). Thus, the chances are relatively high for such an off-target activity of a drug on immune receptors.

Two possibilities have been described for a p-i-mediated drug interaction with the TCR (p-i TCR) or with the HLA molecule (p-i HLA) (Boxes 1, 2) [7]. Docking studies suggest that this off-target drug binding to immune receptors may frequently occur on irrelevant sites of the protein (HLA or TCR) and/or are too labile to elicit a functional consequence [23]. Only occasionally may the drug binding target a critical region of the immune receptor (e.g. peptide binding site of HLA proteins or the CDR3 site of TCR) with a certain affinity and “correct” (stimulatory) orientation of the drug to elicit a profound T-cell activation [7,23,24].

The interaction of a drug with HLA or TCR is often selective for a particular HLA molecule or a particular TCR, as only certain amino-acid sequences and 3D structures allow relatively strong, noncovalent drug binding [7,23,24]. This explains the idiosyncratic nature of p-i-based DHR. It occurs only in some individuals, and persons at risk can be identified by carrying the risk allele.

In contrast to usual off-target activities, where the cell which expresses the off-target receptor is reacting, in p-i exclusively T cells are responsible for the symptoms, even if the HLA molecule of tissue cells is targeted by the drug. This explains the delayed appearance of symptoms mediated by a p-i stimulation, as the initial amount of drug-reactive T cells is often limited and symptoms appear only after T-cell expansion and migration into tissues (similar to hapten-dependent T-cell reactions).

The induction of a strong T-cell reaction, but not drug-induced B-cell stimulations (IgE and IgG), is a characteristic feature of p-i stimulations. It distinguishes p-i from hapten-protein-driven “allergic” DHR, which induce a “complete” T- and B-cell-mediated immune response, with hapten-protein and hapten-peptide specificity (Table 4). Importantly and in spite of the “abnormal,” namely pharmacological T-cell stimulation by p-i, the resulting T-cell activation leads to the secretion of typical cytokines and/or target cell-directed cytotoxicity. Clinical symptoms typically appear >5-7 days after the initiation of treatment. In vitro analysis of T cells of patients suggests that p-i reactions are involved in maculopapular eruptions (MPE) [30], acute generalized exanthematous pustulosis [31], drug-induced liver injury [32], Stevens-Johnson syndrome (SJS)/toxic epidermal necrolysis (TEN) [33,34], and drug reaction with eosinophilia and systemic symptoms (DRESS) [35,36,37,38,39]. Typical examples of drugs known to stimulate via a p-i mechanism are shown in Table 3.

Some drugs may induce immune reactions either by hapten/allergic or by a p-i mechanism, resulting in distinct clinical manifestations. An interesting example is flucloxacillin. Similar to other β-lactam antibiotics, it can act as a hapten (Table 3). However, flucloxacillin-induced liver injury was independent of hapten-linked features of flucloxacillin and showed noncovalent (direct, p-i) binding to HLA B*57:01 [32]. Recent data by D. Yerly et al. [in preparation] show that amoxicillin- and cephalosporin-linked DRESS are also due to p-i stimulations.

There is another, in itself also quite heterogeneous, form of ADR, which was previously referred to as nonallergic DH, drug intolerance; or pseudo-allergy [12,40]. We prefer the term pseudo-allergy, as it emphasizes the close relationship and similarity to allergic symptoms without an underlying drug-specific sensitization [8].

The pathomechanism of pseudo-allergic reactions is not yet completely clarified and is most likely heterogeneous. The symptoms of pseudoallergy can be best explained by the hypothesis that certain drugs activate the effector cells of allergic inflammation, namely mast cells, but also eosinophils, basophils, and probably also neutrophils without evidence of drug-specific IgE, IgG or T cells. We are aware that pseudo-allergic reactions are actually not immune stimulations, and their classification as DHR is thus debatable. But as the clinical picture of pseudo-allergy is so similar to allergic reactions, it makes sense to discuss it together with the other forms of DHR involving the adaptive immune system.

Pseudo-allergic reactions are at least as frequent as IgE-mediated reactions to drugs. Common elicitors of pseudo-allergic reactions are NSAIDs. These drugs act by inhibiting cyclooxygenases and prostaglandin synthesis, resulting in a shift in the arachidonic acid metabolism to leukotriene synthesis [9,41]. In patients with an ongoing, often eosinophilic inflammation like chronic rhinosinusitis, chronic asthma, and chronic spontaneous urticaria, NSAIDs may boost the already existing local inflammation and cause exacerbations of asthma symptoms, rhinosinusitis, and also of urticaria, where inflammatory cells are also present in the skin [42]. Patients with severe asthma have a higher risk to develop NSAID intolerance than individuals with mild or no asthma, and patients with chronic spontaneous urticaria may experience a bout of their urticaria after the intake of NSAIDs [9,42,43,44,45,46]. Thus, the inflammation per se represents the relevant cofactor for the clinical manifestation of NSAID-related pseudo-allergy and explains the localization of symptoms to the inflamed tissue (respiratory tract and/or skin). The intensity of inflammation may even be decisive for the clinical manifestation as the reduction in tissue inflammation may even eliminate the pseudo-allergic symptoms after NSAID intake [45,46].

Many other drugs (e.g. radiocontrast media, muscle relaxants, quinolones, and vancomycin), which have no pharmacological action on prostaglandin synthesis, may also cause pseudo-allergic reactions, in particular acute urticaria and anaphylaxis [8,47]. Also, NSAID may cause acute urticaria and anaphylaxis independent of a preexisting inflammation [42]. Some of these side effects appear to be linked to a single receptor on mast cells, known as MRGPRX2 (Mas-related G-protein-coupled receptor member X2) in humans. This receptor which is recognizing a certain chemical motif was found to be crucial for IgE-independent, direct mast cell stimulation [8]. Fluoroquinolone antibiotics and neuromuscular blocking agents are known to bind to this receptor and to cause systemic pseudo-allergic reactions.

The majority of pseudo-allergic reactions are mild (acute urticaria), but some cause anaphylaxis and can even be lethal, sometimes at the first encounter with the drug as no sensitization is required. Pseudo-allergy is - like IgE-mediated drug allergy - classified as an immediate hypersensitivity reaction, but the mechanism, cross-reactivity, and prevention strategies differ substantially from true allergic reactions (Table 4). Pseudo-allergic reactions are elicited by a limited number of drugs (Table 3), and they are clearly dose dependent. Indeed, they usually require substantially higher doses to elicit clinical symptoms than IgE-mediated reactions, which may occur with the same drugs (e.g. metamizole; Table 3).

Pseudo-allergic reactions are still not predictable, but the detection of relevant and functional off-target receptor(s) on inflammatory cells may allow a better screening of drugs for such an off-target activity. In NSAID-exacerbated respiratory diseases, some links to interleukin-4 promoter polymorphisms and N-acetyl transferase-2 polymorphisms have been described in Korean patients [48,49].

The pharmaceutical industry and regulatory agencies are familiar with the hapten theory of immune-mediated side effects and try to avoid the development of drugs which might act as haptens and modify proteins. Their risk to elicit immune-mediated reactions is considered too high. Indeed, β-lactam antibiotics, which frequently act as haptens, are still the leading cause of drug allergy in daily clinical practice (mainly MPE, urticaria, and anaphylaxis).

However, severe T-cell-mediated DHR appear to be mainly due to a p-i mechanism (see below). Thus, it would be important to avoid such reactions. Screening of novel drugs for their ability to bind to HLA alleles or certain constant regions of TCR, e.g. binding to a TCR-Vβ20.1 [7,24], would be one approach. As the HLA polymorphism is immense, one might first use computational modelling and in silico docking studies. If a strong interaction of a drug with a certain HLA allele at a crucial position is observed, it may be a sign of a relevant off-target activity on the HLA molecule [36]. This needs to be verified by in vitro studies with human cells or possibly humanized animal models to prove the assumed immune-stimulatory potential on human immune receptors. As shown, the orientation of the drug binding to its off-target immune receptor as well as cofactors (e.g. generalized immune stimulations by viral infections) are often crucial to effectively elicit clinical reactions [7,23]. Thus, in silico studies alone are insufficient to predict the hazard. Importantly, in vitro stimulation of drug-specific responses from drug-naïve donors was proven to be possible in quite a number of drugs already if one considered the relevant HLA allele [26,32,36,50]. Thus, a combined computational and biological approach may help to better predict severe, p-i-mediated DHR.

A further rationale for the proposed DHR subclassification is an optimized clinical management: allergic-immune, p-i, and pseudo-allergic reactions lead to distinct recommendations for the affected patient as they differ in relation to drug dose, prediction/HLA linkage, and cross-reactivity (Table 4; Box 3).

Type B reactions are labelled as dose in dependent. This postulation results from observations of rapid anaphylaxis after minimal drug doses, e.g. immediately after the start of an infusion or after skin testing. As the concentrations applied were substantially below the dose needed for a pharmacological effect, the reaction was termed “dose independent.”

In immune reactions, one has to differentiate between the antigen concentration required for the initiation (sensitization) and for the effector phase. This distinction is also relevant in allergic and p-i reactions: the concentrations needed for the initiation (sensitization/cell expansion) are relatively high compared to the concentrations needed for eliciting effector mechanisms and clinical symptoms.

In allergic-immune DHR, the sensitization takes time (>4 days) and occurs to therapeutic drug concentrations. During this time span, the reactive cells expand and mature to effector cells: in antibody-mediated reactions, the B cells generating the antibodies to the hapten-protein complex differentiate and undergo affinity maturation. Interestingly, in IgE-mediated reactions, no symptoms appear during the sensitization phase. In T-cell-mediated reactions, however, symptoms may appear directly following sensitization, namely when the amount of reactive T cells is high enough and homes to the affected organs (mainly the skin).

In already sensitized individuals (having drug-affine IgE or memory T cells), especially on repeated encounter, the reaction appears immediately (IgE, minutes) or rapidly (T cells, hours or a few days) after reexposure and can be elicited by doses which are substantially lower than the ones needed for initiation (sensitization). However, the fact that IgE-mediated reactions can occur already at minimal doses does not mean the reaction is dose independent. Further reducing the dose eliminates symptoms. This allows for drug desensitization procedures [51], where transient tolerance to a drug is induced by initially applying extremely low amounts, which are then steadily increased. After hours to days, even a full drug dose is usually tolerated.

In p-i reactions, drug concentrations are important for eliciting T-cell reactions: symptoms often appear after 7-10 days, the critical time needed for T-cell expansion and migration to the tissue [14]. In DRESS, symptoms may occur even later (>17 days) and rather frequently after further increasing the drug dose or when drug elimination is impaired (renal insufficiency) as well as when drug metabolism is influenced. In allopurinol hypersensitivity, kidney disease with reduced allopurinol clearance and increased allopurinol (oxypurinol) plasma levels is a known cofactor for the development of SJS [52]. In phenytoin hypersensitivity, the presence of the CYP2C9*3 variant, known to reduce phenytoin clearance, was identified as an important genetic factor associated with phenytoin-related severe cutaneous adverse reactions [53].

On the other hand, in highly sensitized individuals (with DRESS or massive MPE), even a minimal drug amount applied for example by intradermal skin testing with the incriminated drug may elicit generalized symptoms like a mild rash (own observation). Thus, lower amounts of the drug may be sufficient to cause symptoms in p-i reactions if a massive expansion of drug-reactive T cells has already taken place.

Pseudo-allergic reactions do not require prior sensitization or cell expansion. Symptoms can appear after the first dose. The clinical symptoms of urticaria or anaphylaxis usually appear at standard to high doses, which is an important distinction to sensitized individuals having IgE reactions. The dose dependence is also well documented by successful desensitization procedures in various pseudo-allergic reactions [54,55]. In NSAID-related pseudo-allergy, the underlying inflammation and effector cell hyperreactivity influences the clinical severity [56].

In conclusion, drug concentrations needed for initiation and elicitation of DHR may vary substantially and differ for allergic, p-i, or pseudo-allergic reactions. Reducing the dose is a useful (although not always sufficient) advice to avoid or minimize DHR.

At present, allergic reactions are not predictable, as they can bind to various epitopes (peptide sequences), which are then presented by different HLA alleles [22]. However, if a prior metabolism of a prohapten to a hapten is involved, differences in the metabolizing enzymes may influence the manifestation of DHR [52,53].

In p-i stimulations, the pharmacological activity of the drug on immune receptors is often linked to the presence of a certain immune receptor in the individual: HLA associations of certain DHR are described for a growing number of drugs [25], some of them with an almost exclusive association to a certain HLA allele, providing a negative predictive value close to 100%, while other drugs were associated with various alleles. On the other hand, the positive predictive value was mostly low (<3%), with the notable exception of abacavir, where about 50% of HLA-B*57:01 carriers developed a hypersensitivity reaction upon drug exposure [57]. The majority of high-risk alleles were HLA class I, but some less stringent associations were also found for HLA class II alleles [58,59].

Predictive testing to avoid DHR is still limited to very few drugs [57,60]. A main problem is the low positive predictive value, suggesting that other factors also play a role for disease manifestation. Such factors may be cytochrome P₄₅₀-related genes [53] or the presence of certain TCR, shown for HLA-B*15:02-related severe carbamazepine hypersensitivity reactions [61]. Thus, a combined genetic analysis of HLA and/or metabolizing enzymes may render p-i-elicited DHR predictable and therefore avoidable.

Cross-reactivity within DHR is a common problem. For allergic DHR, it is usually explained by the affinity of specific immune receptors (IgE antibodies, TCR) for the eliciting drug (hapten-protein or hapten-peptide complexes) but also related structures. The extent of cross-reactivity is dependent on structural similarity and affinity of the drugs to the available immune receptors (TCR, HLA, IgE).

In most instances, the cross-reactive compound binds with lower affinity to the specific immune receptors than the original trigger. This would imply that a higher concentration of the cross-reactive compound is needed to elicit comparable clinical symptoms. Consequently, no symptoms may appear to the cross-reactive compound in T-cell reactions, as the cross-reactive, less-affine compound is not reaching the required dose, which is often similar to the standard dose. In contrast, in IgE-mediated reactions, already very low doses of the original drug are sufficient to elicit symptoms, and a cross-reactive, low-affine compound may still be stimulatory, as the dose given is still in excess.

In p-i reactions, cross-reactivity is based on pharmacology. In the sulfamethoxazole model of p-i TCR, the TCR (“H13”) bound only 6 or all 12 structurally related sulfanilamides at the relevant binding site. In TCR “1.3,” 12 structurally related sulfanilamides bound to the same binding site, but only 1 was functionally active (stimulatory); the orientation of drug binding was crucial for functional activity [7,23,24]. In p-i HLA, abacavir shows a rather restricted ability to bind exclusively to B*57:01, but not to structurally similar HLA-proteins, and modification of the molecule may already abrogate drug binding [62]. In contrast, the carbamazepine-binding HLA-B*15:02 protein binds carbamazepine, some carbamazepine metabolites, and possibly even other anticonvulsants like lamotrigine and phenytoin [63]. Thus, in p-i reactions, cross-reactivity is common, may concern binding to TCR or HLA, and needs to be investigated for each group of compounds separately.

In pseudo-allergy, the secretagogue receptor MRGPRX2 on mast cells binds a structurally heterogeneous group of drugs based on a common chemical motif and thus shows a wide cross-reactivity which cannot be foreseen by structural data [8]. Similarly, in NSAID-induced pseudo-allergy, structurally different NSAID can lead to exacerbations of the disease - as they all have a common mode of action [9,43,44]. This needs to be considered in patient advice and prevention of further reactions.

In conclusion, cross-reactivity in the frame of allergic, p-i, and pseudo-allergic DHR is common, and seems to be most problematic when low drug doses are sufficient to elicit symptoms (IgE-mediated allergy).

An interesting issue is the pathomechanism of SJS/TEN and DRESS. We propose to categorize SJS/TEN and DRESS (including abacavir hypersensitivity syndrome) as typical p-i reactions. This is based on the following arguments:

• When the mode of action of various drugs was analyzed in severe DHR (SJS, DRESS), the stimulation was consistently found to be p-i mediated [32,33,34,35,36,37,38,39]. While this does not rule out that regular T-cell stimulations via an allergic-immune (hapten) mechanism can also lead to such reactions, this seems to be rare.

• Many severe DHR were linked to the presence of a certain HLA allele due to direct drug binding to a critical region of the HLA protein and thus represent a typical “p-i HLA”-type reaction [25,27,28,29,32,33,34,35,36].

• While many of the SJS/TEN/DRESS-causing drugs were shown to stimulate T cells via p-i, B-cell stimulations were not found, and convincing reports on IgE-mediated anaphylaxis to allopurinol, carbamazepine, phenytoin, or abacavir are lacking. An exception are sulfanilamides, where DRESS/SJS and anaphylaxis are known to occur. But sulfanilamide-induced anaphylaxis to SMX is probably elicited by the metabolite SMX-NO, which acts as a hapten, while SJS/DRESS is due to p-i reactions to the parent compound (p-i TCR) [17,23,24].

• The clinical picture of DRESS or SJS/TEN with severe blistering of the skin or a fulminant, possibly lethal hepatitis, carditis, or colitis, for example, after the intake of a small chemical/drug is difficult to comprehend. From a biological point of view such a “suicidal” immune reaction does not make sense. It has been shown that p-i stimulations bypass important control mechanisms inherent in the development of a normal immune reaction [7] (Boxes 1, 2). p-i stimulations can occur without prior activation of dendritic cells or in the absence of danger signals, and naïve and memory T cells are directly activated (shown for abacavir) [26]. Such T-cell stimulations are reminiscent of allostimulations. Indeed, the drug binding to a self HLA molecule may render it to look like an allo-HLA molecule, which elicits a strong allostimulation by T cells [26]. Thus the direct, “at random,” but still affine interaction of drugs with HLA or TCR molecules can lead to a situation comparable to an acute-graft-versus-host reaction [7,26]. Apart from DHR, only graft-versus-host disease is a well-documented cause of SJS/TEN and DRESS [64,65,66]. Thus, severe DHR like SJS/TEN or DRESS may represent a self-destructive reaction against their own cells due to drug modification of their own HLA.

Already in 2003, Aronson and Ferner [67 ]proposed a classification system of ADR (DoTS), which combined properties of the drug and its dose with properties of the reaction (the time course of its appearance and its severity) and of the individual (the genetic, pathological, and other biological differences that confer susceptibility). Here we propose a similar approach to dissect DHR, namely to combine different criteria to define allergic, p-i, and pseudo-allergic DHR subforms (Table 4; Fig. 1).

These criteria are:

• Timing of appearance of symptoms: <1 h (IgE), <1-6 h (pseudo-allergic), or >6 h (delayed/T cells).

• Reaction type (Coombs and Gell) and severity (clinical, laboratory tests) of symptoms (e.g. urticaria, anaphylaxis, mild or bullous exanthema, or involvement of internal organs).

• Known chemical property of the drug (namely to act as hapten/prohapten, via p-i or causing pseudo-allergic reactions like NSAID) and prior experience with the drug, namely what types of DHR are known.

• Skin and in vitro tests: immediate- and delayed-appearing skin tests as well as in vitro tests to define the type of immune mechanism and to pinpoint the eliciting drug and to define safe alternatives.

Timing and type of reaction/symptoms is usually not sufficient to dissect DHR, e.g. rapidly appearing urticaria can be due to allergy (IgE) and/or pseudo-allergy, whereas delayed-appearing exanthems may involve T-cell stimulations by allergic-immune or p-i mechanisms. Thus, additional information like known chemical properties of the drug, experience with the drug, and skin and in vitro tests complete the picture (Boxes 3, 4).

Here we show that immune-mediated DHR develop based on three clearly distinct mechanisms: allergic-immune, p-i, and pseudo-allergic. Only allergic-immune DHR are due to the formation of a new antigen. p-i and pseudo-allergic reactions are pharmacological effects. While allergic and pseudo-allergic reactions have been recognized for many years now, the incorporation of p-i reactions as a distinct subform of DHR is novel. We consider it justified, as p-i-mediated DHR are clearly distinct from allergic or pseudo-allergic reactions and follow different rules. They are not antigen-driven events and are exclusively T-cell mediated. They can be linked to certain HLA alleles or TCR-Vβ subgroups [24]. Importantly, they appear responsible for the most severe T-cell reactions (SJS/TEN/DRESS), but they are not involved in anaphylaxis. p-i-induced symptoms may depend on the presence of cofactors, and they do not need prior exposure, although a cell expansion is often required. They may also differ in desensitization, but this is still unclear. As subclassifying DHR as allergic, p-i, or pseudo-allergic has important implications for patient management and counseling, it is also clinically relevant (Table 4).

In daily clinical practice, the DHR subforms can be distinguished by a combined approach using (a) timing, (b) clinical phenotype/possible mechanisms of immune reactions, and (c) properties and experience with the drug (Table 4; Fig. 1). In addition, immunological tests (skin and in vitro tests) may be required to define the eliciting drug and to demonstrate involvement of specific immune reactions. This combined approach is already practiced in most drug allergy clinics to distinguish allergic from pseudo-allergic reactions. It can be easily extended to consider p-i-mediated reactions as well. Clues for a possible p-i reaction would be (a) the type of the reaction, namely severe forms like SJS/TEN/DRESS, as these are mainly p-i mediated; (b) the property of the drug and history of prior DHR; an exanthema, for example, which is triggered by a known SJS/TEN/DRESS-eliciting drug like allopurinol should be judged differently than an exanthem due to a β-lactam; and (c) a known HLA association which is usually linked to p-i [25] (Fig. 1).

In view of the importance of p-i-mediated DHR for severe reactions, the pharmaceutical industry is challenged to develop tests to identify the risk for such reactions early in drug development. This may actually be advantageous, as p-i reactions may be linked to certain HLA alleles and are thus (partly) predictable [57,60]. The identification of a risk allele may not terminate the development of the incriminated drug, it just limits its use by only excluding carriers of the risk allele. Overall, the drug application might therefore become even safer.

Last but not least, let us learn from our mistakes. The prior classification of DHR as type B reaction and thus “non-A” reaction is too restricted. Statements like “type B reactions are not pharmacologically mediated,” “are not predictable,” or “are dose independent” are misleading, as they do not apply to DHR in general. These statements can be challenged in some definable subforms of DHR, which have significant clinical consequences regarding dose adaptation or predictability (HLA screening).

In conclusion, one can still distinguish ADR type A and type B reactions. Type B reactions are ADR where the immune/inflammatory system is involved. Thus, ADR type B = DHR. DHR have to be further subclassified into allergic-immune, p-i, and pseudo-allergic reactions, which differ in various clinically relevant parameters (Fig. 2).

The support of ADR-AC GmbH is appreciated.

The authors have no conflict of interests to declare.